Marketing is shifting from campaigns to systems. In this interview, Monks co-founder Wesley ter Haar explains why marketing is becoming infrastructure and how agentic workflows are changing execution inside large organizations.
1. Monks chose to place its primary presence at CES Foundry alongside AI infrastructure and hardware companies, rather than remaining solely in the traditional marketing and media space. Is this mainly a signaling move, or does it reflect a deeper shift in who actually controls marketing budgets inside large enterprises today? Are CIOs and CTOs now as central to your buyer conversations as CMOs?
Wesley ter Haar: More and more our focus is on delivering the system (of record) of marketing, and not merely providing the services. That means our clients are predominantly CMO’s, but they tend to be the ones that align closely with the CIO and CTO to make real change happen, beyond the party and parlour tricks of marketing and focused on deep integration of intelligence into the infrastructure of their organization.
It also speaks to a larger trend we see happening because of AI, which is the converging of roles. The CMO, CTO and CIO are solving for the same business outcomes: margin, speed, and real-time relevance which means the CIO has become a central stakeholder for us as we shift from selling isolated campaigns to providing a comprehensive operating system for marketing.
Delivering five times the content on a flat budget is an operational and integration challenge as much as a creative one. We speak the language of infrastructure because that is where the modern brand is built.
2. You’ve described 2026 as the moment when the industry moves from generative AI novelty to discernment, with agentic workflows becoming production-ready. In Monks.Flow, where is the hard boundary between human orchestration and machine autonomy, particularly in high-risk areas like brand strategy, compliance, and reputational judgment?
Wesley ter Haar: The boundary is becoming a fluid exchange between human intent and machine intelligence rather than a static line. While we prioritize having humans in the loop, we are moving beyond simple human-to-agent prompting. The real leverage in 2026 comes from agents prompting people to act based on insights that humans are physically incapable of gathering at scale.
Models possess a distinct advantage in their ability to ingest data in real-time and reason through it across massive datasets. In Monks.Flow, the system acts as a constant cultural sensor, monitoring online conversations, customer reviews, and competitor analysis to identify opportunities as they emerge. When the system recognizes a significant shift, it prompts the human orchestrator with an actionable brief, saying, “there is an opportunity here for the brand to act.”
In high-risk areas like brand strategy and compliance, the hard boundary lies at QA checkpoints and the point of final validation. The AI identifies the signal and crafts the response grounded in the brand’s proprietary DNA, but the human orchestrator remains the final arbiter of reputation and emotional nuance. We use the machine to achieve a level of situational awareness that was previously impossible, while keeping human judgment at the center of the actual execution.
3. Many tools today label themselves “agents” but function more like advanced copilots. What specific capabilities or decision rights must an AI system possess, in your definition, before you consider it truly agentic rather than assistive?
Wesley ter Haar: The distinction lies in orchestration versus assistance. A copilot waits for a prompt; an agent functions as part of an interconnected operating system. To be truly agentic, a system must have the decision rights to move data seamlessly from one stage of a workflow to the next without constant manual intervention.
An agentic system is one where the data flows automatically through repeatable, industrial-grade workflows—from intelligence gathering to creation to precision QA. It isn’t just about generating a response, but rather the system’s ability to reason through a complex chain of tasks. When the AI moves from helping you write an email to autonomously monitoring cultural signals and drafting a campaign for approval, it has crossed the line from assistive to agentic.
4. Monks.Flow is positioned as a model-agnostic orchestration layer, integrating both closed models and open-source models with strong reasoning capabilities. For enterprise clients with strict requirements around data sovereignty, indemnification, and regulatory compliance, how expensive is true model agnosticism in practice, and where does it begin to break down?
Wesley ter Haar: As brands start to truly scale up token usage, the choice and cost of models becomes a financial necessity. Deciding which tokens are generated (potentially on prem) with tuned open source models, versus which are requested from SOTA models via API, will be an ongoing balancing act that will show up at the P&L level.
Monks.Flow enables that flexibility while still having preferred workflow and model combinations ready to go out of the box, based on our own experiences and the expertise of our teams. And all if it is grounded in a data strategy that makes the intelligence proprietary to your brand, while being model agnostic.
5. As Chief AI and Revenue Officer, what now drives growth at Monks: output volume, outcome-based pricing, platform subscriptions, or something else?
Wesley ter Haar: The traditional time-and-materials model is effectively dead in an AI-native world. We have examples where work that used to take weeks and even months, happens close to real-time, although it requires a systems mindset. Growth at Monks is now driven by a shift toward a subscription-based model, we’re combining outcomes and outputs for our clients, delivered by a combination of Talent & Machines. We call this the new T&M model.
It’s faster, better and cheaper, but more importantly, it opens up the right question. We’re no longer asking “what can we afford to do”, we’re focused on “what should we be doing”. The ability to cut 50%+ of costs in the speed & scale parts of the digital ad funnel means AI enables a return to marketing’s original intent, to grow the business by strategically deploying those funds.
6. Monks has described its shift toward a subscription-based innovation model as a replacement for time-and-materials billing. At what point does this model become financially risky for the agency itself, and how confident are you that AI-driven delivery quality is now stable enough to underwrite that risk?
Wesley ter Haar: We see the bigger risk in failing to adapt. Running a commercial model that is based on the time of people has always been a perverse incentive to truly innovate. In the current landscape it will feel completely outdated within years.
When it comes to quality, there are many parts of the marketing supply chain where the current models and agentic workflows compare with or even outcompete purely human efforts. By running the correct “people in the lead / people in the loop” workflows we get the best of both worlds: the scale, speed and efficiencies of systems at scale, with the taste and discernment of deep subject matter experts.
7. You often talk about moving talent from operators to orchestrators. Historically, creative judgment was forged through years of hands-on execution. If AI absorbs most of that execution layer, how does Monks prevent a future generation of leaders from losing that instinctive sense of craft and detail?
Wesley ter Haar: This is a common fear, but I see it as a democratization of craft. Historically, “craft” was often synonymous with drudgery: the manual resizing of 50 assets or the technical minutiae of VFX, for example. By removing that heavy lifting, we are actually allowing our talent to focus more on the high-level vision and narrative.
The instinctive sense of craft now evolves into the ability to tune a model and provide high-fidelity reinforcement feedback. Our talent is moving below the fold; they’re building the knowledge bases and the prompts that define the brand’s soul. The best users of these tools are still the ones with strong creative perspective and expertise. We aren’t losing craft; we’re elevating it from pixel-pushing to world-building.
8. The shift from demographic targeting to real-time “culture graphics” promises extreme relevance, but it also encourages constant adaptation to short-lived cultural signals. How do you prevent brands from becoming reactive and fragmented, optimizing for immediacy at the expense of long-term identity?
Wesley ter Haar: Culture graphics allow brands to move beyond mere reaction and become culturally indispensable. While traditional demographics categorize people by who they are, culture graphics focus on what they care about in the moment.
We prevent fragmentation by grounding every real-time action in the brand’s Intelligence. Every cultural response is filtered through a proprietary knowledge base that understands the brand’s long-term DNA. We use AI to listen to the world at a scale humans can’t, but that listening is always calibrated against the brand’s core identity. It’s about being “liquid” in execution while remaining “solid” in strategy.
9. LiveVision’s use of edge computing and real-time AI inference places Monks closer to broadcast infrastructure and systems integration than traditional marketing services. Does this signal that Monks’ long-term moat lies less in creativity and more in technical integration of physical and digital systems? How do you compete with consultancies that already dominate enterprise IT implementation?
Wesley ter Haar: Our moat is the integration of creativity and technology, as evidenced by the way we structure our business: marketing services and technology services.
LiveVision is a perfect example of a workflow that enables you to build content that adapts as fast as the audience does. We compete with consultancies by being faster and more culturally attuned, and instead of just building the pipes, we know exactly what should flow through them to make a brand culturally relevant. We aim to be the strategic experts in how technology adapts to, reinforces, and influences human behavior.
10. Persona.Flow allows brands to test ideas against synthetic consumers at near-zero marginal cost. How do you guard against AI systems validating ideas that appear sound in simulation but fail in real human, emotional, and irrational market conditions?
Wesley ter Haar: This concern often assumes that traditional research methods—like focus groups or self-reported surveys—are a perfect baseline for truth. In reality, humans are notoriously poor at predicting their own irrational behavior in a room full of strangers. We’ve been building brands based on imperfect, biased human data for decades.
Persona.Flow isn’t a replacement for human reality, but a high-fidelity tool for removing friction and validating the logic of an idea before it ever reaches a human. Our guardrails are built into our intelligence: we ground synthetic personas in deep, third-party data and real-world cultural insights rather than generic model responses. This ensures the simulation is as representative of human behavior as possible.
The final guardrail remains the human orchestrator. We use Persona.Flow to get us to the starting line of a great idea at near-zero cost, but we keep a human in the lead to validate the emotional spark. We are optimizing for speed and scale in the iterative phase so that when we do engage with real humans, we are doing so with an idea that has already been stress-tested against the brand’s DNA and broad cultural signals.
11. Publicis has built CoreAI around a massive identity graph, while Monks has focused on speed, orchestration, and content production rather than proprietary consumer identity. In a privacy-constrained, post-cookie world, is the lack of a first-party identity asset a structural weakness for Monks, or a deliberate strategic tradeoff?
Wesley ter Haar: We do have a focus on consumer identity—we just believe the brand should own it, not the agency. The massive identity graphs of the past were often built on third-party data that is becoming increasingly unstable and non-compliant in a privacy-first world.
So, helping brands build and leverage their own first-party data is an important focus area for us. Whether it’s through our CRM practices that drive one-to-one customer relationships or our privacy-first, first-party data assessments, we are building durable, scalable foundations that a brand actually owns. We empower brands to wield their own data as a strategic asset.
This foundational data is the essential fuel for everything we build; you cannot have high-fidelity AI orchestration without a deep, clean understanding of the customer journey. Culture graphics serve as a layer of real-time relevance that makes a brand’s 1P identity actionable. Instead of chasing individuals through rented graphs, we connect customer experiences across digital contexts, ensuring brands remain data rich in the moments that matter.
12. Looking toward 2027, what do you see as the hardest unresolved bottleneck preventing a truly AI-native marketing organization: model reliability, inference cost, long-term memory, organizational trust in machine decisions, or something more human that technology alone cannot solve?
Wesley ter Haar: Technology is no longer the blocker; it’s us. The biggest bottleneck is human and organizational change management.
We have the technology to do things in two weeks that used to take nine months, but most organizations aren’t built to ingest work at that speed. Our commercial models, our internal silos, and our legacy waterfall workflows are the real barriers. To become truly AI-native, companies have to move from a playground mindset to a production one. 2027 won’t be about who has the best model; it will be about who has the most resilient and adaptive organization.
Editor’s Note
This interview frames marketing as infrastructure rather than a creative function. As AI-native systems mature, execution speed and integration move from optimization goals to organizational requirements.
